NVIDIA Unveils Vera Rubin POD Cluster: Five Racks, Seven Chips, 1.2 Quadrillion Transistors

NVIDIA unveiled the Vera Rubin POD, a cluster consisting of five specialized rack systems based on the third-generation MGX architecture. It includes NVL72 racks (with 72 Rubin accelerators and 36 Vera processors), LPX racks (256 LPU accelerators for low latency), Vera racks (up to 256 processors for agentic AI), STX racks (storage based on BlueField-4 for contextual memory), and SPX racks (networking racks with Spectrum-X Ethernet or Quantum-X800 InfiniBand). The cluster comprises 40 racks, 1.2 quadrillion transistors, nearly 20,000 NVIDIA dies, and 1,152 Rubin GPUs. Performance reaches 60 exaflops, with aggregated interconnect speed of 10 petabytes per second. MGX racks feature unified power and cooling parameters, dynamic power management, and intelligent power consumption smoothing. Cooling uses hot water up to +45°C. Shipments of Vera Rubin NVL72 will begin in the second half of 2026.